Slice Profile Manager: O-RAN-Based Optimization of Slice Profiles using Reinforcement Learning

With the progression of 5G networks, intelligent resource management becomes crucial, especially in network slicing, where efficiency and flexibility are essential. This paper presents a novel architecture that integrates a Deep Q-Network (DQN) powered Slice Profile Manager (SPM) with the O-RAN Alliance (O-RAN) framework. The SPM dynamically optimizes slice profiles in real-time using reinforcement learning to adjust resource allocation across Radio Access Network (RAN), Transport Network (TN), and Core Network (CN) based on changing network conditions. This research aims to improve traditional network slicing methodologies by facilitating enhanced situational awareness and autonomous decision-making capabilities. The study introduces a real-time slice configuration method called SPM, an architectural model based on O-RAN that includes SPM and a mathematical model for DQN-based optimization. This research lays the groundwork for addressing network challenges and enhancing adaptability for more responsive 5G and beyond networks.